KR101762210B1 - Encoding method, encoder, program and recording medium - Google Patents

Abstract

The number of bits or the number of bits to be obtained by coding a column of an integer value sample obtained by dividing each sample of the sample string of the sample string derived from the input acoustic signal of a predetermined section by the gain before the update and the predetermined number of distribution bits B ), The gain value is updated so that the difference between the gain before update and the gain after update becomes larger, and a gain code corresponding to the obtained gain and a column with an integer value sample obtained by dividing each sample of the sample string by the gain are encoded To obtain an integer signal code.

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention [0001] The present invention relates to an encoding method, an encoding apparatus, a program,

TECHNICAL FIELD The present invention relates to a technique for encoding acoustic signals. Particularly, the present invention relates to a sequence encoding technique obtained by dividing a sample sequence derived from an acoustic signal by a gain.

As an encoding method of a speech signal or a sound signal of a low bit (for example, about 10 kbit / s to 20 kbit / s), adaptive encoding for orthogonal transform coefficients such as DFT (Discrete Fourier Transform) or MDCT (Modified Discrete Cosine Transform) Is known. For example, AMR-WB + (Extended Adaptive Multi-Rate Wideband), which is a standard specification of non-patent document 1, has a transform coded excitation (transform coded excitation) coding mode. In TCX encoding, a coefficient sequence obtained by normalizing an acoustic digital signal sequence in the frequency domain by a power spectral envelope coefficient sequence so as to be able to perform coding with a given total number of bits per frame is obtained by dividing each coefficient in the coefficient sequence by a gain. The gain is determined so that the sequence can be encoded with a predetermined number of bits.

<TCX encoding apparatus 1000>

A configuration example of a conventional TCX encoding apparatus 1000 is shown in Fig. Hereinafter, each section of Fig. 1 will be described.

&Lt; Frequency domain converter 1001 >

The frequency domain transforming unit 1001 transforms the inputted acoustic digital signal into N MDCT coefficient columns (X (1), ..., X (N)) in the frequency domain in units of frames which are a predetermined time period Output. Where N is a positive integer.

&Lt; Power spectral envelope coefficient column calculation unit 1002 >

The power spectral envelope coefficient arithmetic unit 1002 performs linear prediction analysis on an acoustic digital signal on a frame-by-frame basis to obtain a linear prediction coefficient, and uses the linear prediction coefficient to compute a power spectral envelope coefficient column of the N acoustic digital signals W (1), ..., W (N)).

<Weighted-envelope normalization unit 1003>

The weighted envelope normalization unit 1003 normalizes each coefficient of the MDCT coefficient sequence obtained by the frequency domain transform unit 1001 using the power spectral envelope coefficient sequence obtained by the power spectrum envelope coefficient column calculation unit 1002, And outputs MDCT coefficient columns X _N (1), ..., X _N (N). In order to realize the quantization in which the deformation of the auditory sound is reduced, the weighted envelope normalization unit 1003 uses the weighted power spectral envelope coefficient column which weakens the power spectrum envelope, and calculates the coefficients of the MDCT coefficient columns Normalize. As a result, the weighted normalized MDCT coefficient columns (X _N (1), ..., X _N (N)) do not have irregularities of inclination or amplitude of a large amplitude about the input MDCT coefficient column, A small amplitude similar to the spectral envelope coefficient column, that is, a region on the coefficient side corresponding to the low frequency, and a fine structure due to the pitch period.

<Initialization Unit 1004>

The initialization unit 1004 sets the initial value of the gain (global gain) g. The initial value of the gain is determined by the energy of the weighted normalized MDCT coefficient column (X _N (1), ..., X _N (N)) and the number of bits previously allocated to the code output from the variable length coding unit 1006 . Hereinafter, the number of bits allocated in advance to the code output from the variable length coding unit 1006 is referred to as a distribution bit number (B). Further, the initialization unit sets 0 as an initial value of the number of times of updating of the gain.

&Lt; Gain Update Loop Processor 1130 >

The gain update loop processing unit 1130 encodes a series obtained by dividing each coefficient in the weighted normalized MDCT coefficient columns X _N (1), ..., X _N (N) by a gain into a predetermined number of bits And obtains a constant signal code obtained by performing variable length coding on a series obtained by dividing the weighted normalized MDCT coefficient string (X _N (1), ..., X _N (N) And a gain code obtained by encoding the determined gain.

The gain update loop processing unit 1130 includes a quantization unit 1005, a variable length coding unit 1006, a determination unit 1007, a gain expansion update unit 1131, a gain reduction update unit 1132, Unit 1016, and a gain encoding unit 1017. [

<Quantization Unit 1005>

The quantization unit 1005 quantizes a value obtained by dividing each coefficient of the weighted normalized MDCT coefficient column X _N (1), ..., X _N (N) by a gain g, and it outputs the obtained quantized normalized coefficient sequence _{(X Q (1), ···} , X Q (N)).

<Variable-length coding unit 1006>

Variable length coding unit 1006 outputs a code obtained by variable length coding the quantized normalized coefficient sequence _{(X Q (1), ···} , X Q (N)). This code is called an integer signal code. In this variable-length coding, for example, a method of arranging a plurality of coefficients in the quantized normalized coefficient sequence and encoding them is used. The variable length coding unit 1006 measures the number of bits of the integer signal code obtained in the variable length coding. Hereinafter, this number of bits is called the number of consumed bits (c).

&Lt; Judgment section 1007 >

When the number of times of updating the gain is a predetermined number, the determining section 1007 outputs the gain, the integer signal code, and the number of consumed bits (c).

If the number of times of updating of the gain is less than the predetermined number, the gain increase updating unit 1131 updates the variable length coding unit 1006 when the number of consumed bits c measured by the variable length coding unit 1006 is larger than the distribution bit number B, When the number of consumed bits c measured by the encoding unit 1006 is smaller than the number of allocated bits B, the gain reduction updating unit 1132 performs control to perform the following processing. When the number of consumed bits c and the number of allocated bits B are equal to each other, it means that the value of the present gain is the optimum value, and outputs the gain, the integer signal code, and the number of consumed bits c.

<Gain Expansion Updating Unit 1131>

The gain increase updating unit 1131 sets a value (g '> g) larger than the current gain g as a new gain. The gain enhancement updating unit 1131 has a gain lower limit setting unit 1008, a first branching unit 1009, a first gain updating unit 1010, and a gain increasing unit 1011.

&Lt; Gain lower limit setting unit 1008 >

Gain lower limit setting unit 1008 sets the value of the gain (g) of this time as a lower limit value (g _min) of the gain (g _min ← g). The lower limit value ( _gmin ) of this gain means that the value of the gain should be at least above this value.

&Lt; First branch portion 1009 >

If the gain lower limit setting unit 1008, the lower limit value (g _min) of a gain is set in the first branch 1009 is a case that the upper limit value (g _max), the gain is already set, the first gain updating unit 1010 , And otherwise controls the gain increasing unit 1011 to perform the next process.

<First Gain Updating Unit 1010>

The first gain updating unit 1010 newly sets the value of the current gain g and the average value of the upper limit value g _max of the gain as the value of the gain g newly (g? G + g _max ) / 2). This is because the value of the optimum gain exists between the value of the current gain g and the upper limit value g _max of the gain. The value of the gain (g) for this time is also that it is set as a lower limit value (g _min) of the gain, the new mean value of the upper limit value (g _max) and the gain lower limit value (g _min) of the gain set to the value of the gain (g) (G? ( _Gmax + _gmin ) / 2). And thereafter returns to the processing of the quantization unit 1005.

<Gain Expansion Unit 1011>

The gain increasing unit 1011 sets a value larger than the value of the current gain g as a value of the new gain g. For example, a value obtained by adding a gain change amount? G, which is a predetermined value to the value of the current gain g, is set as the value of the new gain g (g? G +? G). Further, for example, when the state where the upper limit value g _max of the gain is not set and the number of consumed bits c is more than the distribution bit number B is continued plural times, a value larger than a predetermined value is set to Is used as the gain change amount? G. And thereafter returns to the processing of the quantization unit 1005.

<Gain Reduction Updating Unit 1132>

The gain reduction updating unit 1132 sets a value (g '<g) smaller than the value of the current gain g as a new gain. The gain reduction unit 1132 has a gain upper limit setting unit 1012, a second branching unit 1013, a second gain updating unit 1014, and a gain reducing unit 1015.

<Gain upper limit setting unit 1012>

The gain upper limit setting unit 1012 sets the value of the current gain g to the upper limit value g _max of the gain (g _max? G). The upper limit value ( _gmax ) of this gain means that the value of the gain should be at least below this value.

&Lt; Second branching part 1013 >

If the gain upper limit setting portion 1012 is set the upper limit value (g _max) of the gain, and the second branch 1013 is if the lower limit value (g _min) of a gain is already set, the second gain updating unit 1014, the , And otherwise controls the gain reducing unit 1015 to perform the next processing.

&Lt; Second gain updating unit 1014 >

Second gain updating unit 1014 is, for example, and sets the average value of the gain (g) values and the gain lower limit value (gmin) of the current time as the value of the new gain (g) (g ← (g + g _min )/2). This is because the value of the optimum gain is to exist between the lower limit value (g _min) of a gain value and a gain (g) of the current time. The value of the gain (g) for this time is also that it is set as the upper limit value (g _max) of the gain, the new mean value of the upper limit value (g _max) and the gain lower limit value (g _min) of the gain set to the value of the gain (g) (G? ( _Gmax + _gmin ) / 2). And thereafter returns to the processing of the quantization unit 1005.

&Lt; Gain reducing unit 1015 >

The gain reduction unit 1015 sets a value smaller than the current gain g as a value of the new gain g. For example, a value obtained by subtracting the gain change amount? G, which is a predetermined value from the value of the current gain g, is set as the value of the new gain g (g? G-? G). Also, for example, when no minimum value (g _min) of the gain is not set, the number of spent bits (c) The allocation bit number (B) with less state continues a plurality of times, a value that is greater than a predetermined value Is used as the gain change amount? G. And thereafter returns to the processing of the quantization unit 1005.

<Discard part 1016>

When the number of consumed bits c output from the determining section 1007 is larger than the number of allocated bits B, the discard section 1016 sets the number of consumed bits c (c) out of the integer signal codes output from the determining section 1007 ) Is removed from the code corresponding to the quantized normalized coefficient on the higher frequency side, as a new integer signal code. In other words, the discard unit 1016 obtains the quantization normalized coefficient obtained by removing the sign corresponding to the higher frequency side quantized normalized coefficient corresponding to the upper fraction cB with respect to the distribution bit number B of the consumed bit number c from the integer signal code , And outputs the remaining sign as a new integer signal code.

&Lt; Gain coding unit 1017 >

Obtains a gain code by encoding the gain output from the determination section 1007 to a predetermined number of bits, and outputs the gain code.

 (3GPP), Technical Specification (TS) 26.290, "Extended Adaptive Multi-Rate-Wideband (AMR-WB +) codec; Transcoding functions", Version 10.0.0 (2011-03)

(Summary of the Invention)

(Problems to be Solved by the Invention)

The gain enlarging unit 1011 of the conventional encoding apparatus 1000 fixedly adds the value of the gain g to the value of the gain g by adding the gain change amount? G, which is a predetermined value, .

When the upper limit value of the gain is not set and the process of the gain increasing unit 1011 is required a plurality of times, the initial value of the gain may be too small to the pole, so that the gain changing amount? The probability of reaching the upper limit value of the gain can not be increased. However, a value greatly exceeding the proper gain may be set as the value of the new gain, and the number of times of convergence is required, The value of the appropriate gain can not be obtained by the number of times.

Likewise, in the gain reducing unit 1015 of the conventional encoding apparatus 1000, the value obtained by subtracting the gain change amount? G, which is a predetermined value, from the value of the gain g is set as the value of the new gain g, The gain value was reduced.

If the lower limit value of the gain is not set and the processing of the gain reducing section 1015 is required a plurality of times, the initial value of the gain may be excessively large. Therefore, the gain change amount? G may be made larger than a predetermined value, The probability of reaching the lower limit value of the gain can not be increased. However, a value greatly exceeding the appropriate gain may be set as the value of the new gain, and the number of times of convergence is required. The value of the gain can not be obtained in some cases.

If the value of the gain obtained by the predetermined number of times is too small, the number of bits of the code obtained by the variable length coding is larger than the allocation bit, so that only a part of the code obtained by the variable length coding can not be converted into the integer signal code. Since the code corresponding to the quantized normalized coefficient of the high frequency band is not outputted from the coding device nor transmitted to the decoding device, the decoding device must obtain a decoded signal by setting the coefficient of the high frequency band to 0, There is a problem that the deformation becomes large. If the value of the gain determined by the predetermined number of times is too large, the number of bits of the integer signal code is smaller than the allocation bit, so that there is a problem that sufficient quality of the acoustic signal can not be obtained.

The number of bits or the estimated number of bits and the predetermined number of distribution bits (B) obtained by coding a column of integer value samples obtained by dividing each sample of the sample string of the sample string derived from the input acoustic signal of a predetermined section by the gain before the update, The gain value is updated so that the difference between the gain before the update and the gain after the update becomes larger as the difference between the gain before the update and the gain after the update becomes larger, and the gain code corresponding to the obtained gain and the column with the integer value obtained by dividing each sample of the sample string by the gain To obtain an integer signal code.

According to the encoding method of the present invention, it is possible to bring the number of bits of the code obtained by the variable length coding closer to the distribution bits than that of the prior art, by accelerating the convergence of the gain value to an appropriate value, Can be performed.

1 is a block diagram illustrating a configuration of a conventional encoding apparatus;
2 is a block diagram exemplifying a configuration of an encoding apparatus according to the first embodiment;
3 is a block diagram exemplifying a configuration of a coding apparatus according to a modification of the first embodiment;
4 is a block diagram illustrating the configuration of an encoding apparatus according to a second embodiment;
5 is a block diagram exemplifying a configuration of a coding apparatus according to a modification of the second embodiment;
6 is a block diagram exemplifying a configuration of an encoding apparatus according to a third embodiment;

(Mode for carrying out the invention)

Embodiments of the present invention will be described with reference to the drawings. The same reference numerals are assigned to the same component or the same process, and redundant description may be omitted. In addition, the acoustic digital signal (input acoustic signal) handled in each embodiment is a signal in which acoustic signals such as voice or musical tones are digitized. In each embodiment, the input acoustic digital signal is a time-domain signal of a predetermined time interval, and the acoustic digital signal is converted into a frequency domain signal, and a column obtained by normalizing the frequency domain signal using the power spectrum envelope coefficient string Is a sample string to be encoded (a sample string derived from an input acoustic signal). However, if the input acoustic digital signal is a time-domain signal of a predetermined time interval, the acoustic digital signal itself may be a sample stream to be encoded, and a residual signal obtained by performing linear prediction analysis on the acoustic digital signal may be a sample- Sample stream, and the frequency domain signal converted from the acoustic digital signal may be a sample stream to be encoded. Alternatively, if the inputted acoustic digital signal is a frequency domain signal of a predetermined section (a frequency domain signal corresponding to a predetermined time interval or a frequency domain signal of a predetermined frequency domain of the frequency domain signal) The time-domain signal converted from the acoustic digital signal may be a sample stream to be encoded, and the residual signal obtained by performing a linear prediction analysis on the time-domain signal may be a sample stream to be encoded. That is, the inputted acoustic digital signal may be either a time domain signal or a frequency domain signal, and a sample stream to be subjected to the encoding processing may be a time domain signal or a frequency domain signal. Also, there is no limitation on a method of converting a time domain signal into a frequency domain signal and a method of converting a frequency domain signal into a time domain signal. For example, a modified discrete cosine transform (MDCT) or a discrete cosine transform (DCT) Inverse transformation and the like can be used.

On the basis of the above assumption, in each of the embodiments, the encoder has a frequency domain transformer, a power spectral envelope coefficient column calculator, and a weighted envelope normalizer, and an example in which the sample string obtained in the weighted envelope normalizer is input to the quantizer present. However, in the case where the input acoustic digital signal itself is used as a sample sequence to be encoded, for example, a frequency domain transform unit, a power spectral envelope coefficient column calculating unit, and a weighted envelope normalizing unit are omitted, Is directly input to the quantization unit. When a residual signal obtained by performing a linear prediction analysis on an input acoustic digital signal as a time-domain signal is used as a sample stream to be encoded, for example, the encoding apparatus includes a frequency domain transform unit, a power spectrum envelope coefficient column calculator, A linear prediction unit that receives an acoustic digital signal as input and obtains a linear prediction coefficient or a coefficient convertible thereto, a linear prediction filter that corresponds to the linear prediction coefficient, and a residual calculation unit that obtains a prediction residual from the acoustic digital signal And a sample sequence of the residual signal is input to the quantization unit. For example, the power spectral envelope coefficient column calculator and the weighted envelope normalizer are omitted in the case where the frequency-domain signal converted from the input digital signal as the input time-domain signal is used as the sample string to be encoded, A sample sequence of the obtained frequency domain signal is input to the quantization section. In the case where the time-domain signal converted from the acoustic digital signal, which is the input frequency-domain signal, is used as the sample string to be coded, for example, the coding apparatus includes a frequency domain transformer, a power spectrum envelope coefficient column calculator and a weighted envelope normalization And a time domain converter for converting the acoustic digital signal into a time domain signal instead of the time domain signal. The sample sequence of the time domain signal is input to the quantizer. When a residual signal obtained by performing a linear prediction analysis on a time-domain signal transformed from an acoustic digital signal, which is an input frequency-domain signal, is used as a sample stream to be encoded, for example, A time-domain transform unit, a linear predictor, and a residual calculating unit instead of the envelope coefficient column calculating unit and the weighted envelop normalizing unit, and the sample sequence of the residual signal obtained by the residual calculating unit is input to the quantizing unit.

[First Embodiment]

&Lt; Encoder 100 >

The encoding process performed by the encoding apparatus 100 of the first embodiment will be described with reference to Fig.

<Frequency Domain Conversion Unit 101>

The frequency domain transforming unit 101 transforms the input acoustic digital signal (input acoustic signal) into N MDCT coefficient strings in the frequency domain (X (1), ..., X (N )) And outputs it. However, N is a positive integer.

&Lt; Power spectral envelope coefficient arithmetic unit 102 >

The power spectral envelope coefficient arithmetic unit 102 calculates a linear predictive coefficient by performing a linear prediction analysis on an acoustic digital signal on a frame basis and calculates a power spectral envelope coefficient column of the acoustic digital signal of N points using the linear predictive coefficient W (1), ..., W (N)).

<Weighted-envelope normalization unit 103>

The weighted envelope normalization unit 103 normalizes each coefficient of the MDCT coefficient sequence obtained by the frequency domain transform unit 101 using the power spectral envelope coefficient sequence obtained by the power spectrum envelope coefficient column calculation unit 102 and outputs the weighted normalized MDCT And outputs the coefficient sequence X _N (1), ..., X _N (N). Here, in order to realize quantization in which the deformation becomes small audibly, the weighted envelope normalization unit 103 normalizes each coefficient of the MDCT coefficient column in frame units by using a weighted power spectral envelope coefficient column that weakens the power spectrum envelope do. As a result, the weighted normalized MDCT coefficient columns (X _N (1), ..., X _N (N)) do not have irregularities of inclination or amplitude of a large amplitude about the input MDCT coefficient column, Has a similar magnitude to the spectral envelope coefficient column, that is, has a slightly larger amplitude in the area on the coefficient side corresponding to the lower frequency, and has a fine structure due to the pitch period.

[Specific example of weighted envelope normalization processing]

W (N) of the power spectral envelope coefficient column corresponding to each coefficient (X (1), ..., X (N)) of the MDCT coefficient column of N points is linear Can be obtained by converting the prediction coefficients into the frequency domain. For example, the time signal x (t) at time t can be obtained by a p-th order autoregressive process (where p is a positive integer), which is an electrode model, by x (t-1), ··· , x (tp)) and a prediction residual (e (t)) and a linear predictive coefficient _{(α 1, ···, α p} ) is represented by the formula (1). At this time, each coefficient (W (n)) [1? N? N] of the power spectral envelope coefficient column is expressed by equation (2). exp (·) is an exponential function whose number is lower than the number of Napier, j is an imaginary unit, and σ ² is the predicted residual energy.

(Equation 1)

The linear predictive coefficient may be obtained by linear predictive analysis of the acoustic digital signal input to the frequency domain transform unit 101 by the weighted envelope normalization unit 103 or by other means not shown in the encoding apparatus 100 It may be obtained by linear prediction analysis of an acoustic digital signal. In this case, the weighted envelope normalization section 103 uses the linear prediction coefficients to obtain the respective coefficients W (1), ..., W (N) of the power spectral envelope coefficient row. (W (1), ..., W (N)) of the power spectral envelope coefficient column are already obtained by other means (such as the power spectral envelope coefficient column calculating section 102) in the encoding apparatus 100 The weighted envelope normalization unit 103 can use the respective coefficients W (1), ..., W (N) of this power spectral envelope coefficient column. Also, since it is necessary to obtain the same value as that obtained by the encoding apparatus 100 also in the decoding apparatus, quantized linear prediction coefficients and / or power spectral envelope coefficient columns are used. In the following description, &quot; linear prediction coefficient &quot; or &quot; power spectral envelope coefficient column &quot; means a quantized linear prediction coefficient or power spectral envelope coefficient column unless otherwise specified. Further, the linear prediction coefficients are coded by, for example, a conventional coding technique, and the prediction coefficient codes are transmitted to the decoding side. Conventional encoding techniques include, for example, a coding technique that uses a code corresponding to the linear prediction coefficient itself as a prediction coefficient code, a coding technique that converts a linear prediction coefficient to an LSP parameter and uses a code corresponding to the LSP parameter as a prediction coefficient code A coding technique for converting a linear prediction coefficient into a PARCOR coefficient, and using a code corresponding to the PARCOR coefficient as a prediction coefficient code. In the case where the power spectral envelope coefficient sequence is obtained by other means in the encoding apparatus 100, the linear prediction coefficients are encoded by the conventional encoding technique in other means in the encoding apparatus 100, And transmitted to the decoding side.

Here, two specific examples of the weighted envelope normalization process are presented, but the present invention is not limited to these examples.

<Example 1>

The weighted envelope normalization unit 103 multiplies each coefficient (X (1), ..., X (N)) of the MDCT coefficient column by the correction value W _? (1 ), ···, W _γ (N) by dividing a), the weighted normalized MDCT coefficients of columns for each coefficient _{(X (1) / W γ} (1), ···, X (N) / W γ (N)) Is obtained. The correction value W _? (N) (1 _? N? N) is given by equation (3). However,? Is a positive constant of 1 or less and is a constant that weakens the power spectral coefficient.

(Equation 2)

<Example 2>

The weighted envelope normalization unit 103 multiplies each coefficient (X (1), ..., X (N)) of the MDCT coefficient string by the beta power of each coefficient of the power spectrum envelope coefficient column corresponding to each coefficient (0 < 1) of the value ^{(W (1) β, ···} , W (N) β) by dividing, the weighted normalized MDCT coefficients of columns for each coefficient (X (1) / W ( 1) β, ···, X ( N) / W (N) ^? ).

As a result, a weighted normalized MDCT coefficient column in frame units is obtained. The weighted normalized MDCT coefficient column does not have a large amplitude inclination or amplitude irregularity about the input MDCT coefficient column, but is similar to the power spectrum envelope of the input MDCT coefficient column That is, the area on the coefficient side corresponding to the low frequency has a slightly large amplitude and has a fine structure due to the pitch period.

In addition, since the inverse process corresponding to the weighted envelope normalization process, that is, the process of restoring the MDCT coefficient sequence from the weighted normalized MDCT coefficient sequence is performed on the decoding side, a method of calculating the weighted power spectrum envelope coefficient column from the power spectrum envelope coefficient sequence To be common to the encoding side and the decoding side.

<Initialization Unit 104>

The initialization unit 104 sets the initial value of the gain (global gain) g. The initial value of the gain is determined by the energy of the weighted normalized MDCT coefficient column (X _N (1), ..., X _N (N)) and the number of bits allocated in advance by the code output from the variable length coding unit 106 . The initial value of the gain g is a positive value. Hereinafter, the number of bits allocated in advance to the code output from the variable length coding unit 106 is referred to as a distribution bit number (B). Further, the initialization unit sets 0 as an initial value of the number of times of updating of the gain.

&Lt; Gain Update Loop Processor 130 >

The gain update loop processing unit 130 sets a series obtained by dividing each coefficient in the weighted normalized MDCT coefficient columns X _N (1), ..., X _N (N) by a gain (Integer number of samples) obtained by dividing the coefficient in the weighted normalized MDCT coefficient string (X _N (1), ..., X _N (N)) by the determined gain, , And a gain code (gain code corresponding to gain) obtained by coding the determined gain are output. The gain update loop processing unit 130 sets the difference between the gain before update and the gain after update to be larger as the difference between the number of bits of the code obtained by encoding the column by the integer value samples and the predetermined number of distribution bits (B) And updates the value of the gain.

The gain update loop processing unit 130 includes a quantization unit 105, a variable length coding unit 106, a determination unit 107, a gain expansion update unit 131, a gain reduction update unit 132, Unit 116, and a gain encoding unit 117. [

<Quantization Unit 105>

The quantization unit 105 quantizes each coefficient (each sample) of input weighted normalized MDCT coefficient columns X _N (1), ..., X _N (N) (sample sequences derived from input acoustic signals of a predetermined section) Quantized normalized coefficient series X _Q (1), ..., X _Q (N)), which is a series based on an integer value (a quantized normalized sample) do.

The quantization unit 105 also counts the number of samples (s) from the quantized normalized coefficient on the lowest frequency side to the quantized normalized coefficient on the highest frequency side where the value is not 0, ).

<Variable-length coding unit 106>

The variable length coding unit 106 outputs the inputted quantized normalized coefficient sequence _{(X Q (1), ···} , X Q (N)) to the variable-length encoding takes the sign (sign of samples). This code is called an integer signal code. In this variable-length coding, for example, a method of arranging a plurality of coefficients in the quantized normalized coefficient sequence and encoding them is used. The variable length coding unit 106 measures the number of bits of the integer signal code obtained in the variable length coding. In this embodiment, this number of bits is called the number of consumed bits (c).

&Lt; Judgment section 107 >

The determination unit 107 outputs the gain g, the integer signal code, and the number of consumed bits c when the number of times of updating the gain is a predetermined number.

When the number of times of updating of the gain is less than the predetermined number, the gain enlargement updating unit 131 updates the variable length encoding unit 106 when the number of consumed bits c measured by the variable length coding unit 106 is larger than the distribution bit number B, When the number of consumed bits c measured by the encoding unit 106 is smaller than the number of allocated bits B, the gain reduction updating unit 132 performs control to perform the next process. When the number of consumed bits c measured by the variable length coding unit 106 is equal to the number of allocated bits B, the determining unit 107 determines the gain g, the integer signal code, and the number of consumed bits c ).

<Gain Expansion Updating Unit 131>

The gain increase updating unit 131 sets a value (g '> g) larger than the current gain g as a new gain. The gain enlarging and updating unit 131 includes a sample number measuring unit 118, a gain lower limit setting unit 108, a first branching unit 109, a first gain updating unit 110, a gain enlarging unit 111, .

&Lt; Sample number measuring section 118 >

When the number of consumed bits c is larger than the number of allocated bits B, the number-of-samples measuring unit 118 sets the number of consumed bits c to the number of allocated bits B), the number of samples of the quantized normalized coefficient corresponding to the code from which the sign corresponding to the quantized normalized coefficient on the high frequency side is removed is output.

That is, the number-of-samples measuring unit 118 quantizes the high-frequency-side quantized normalized coefficient corresponding to the sign (decimal notation) corresponding to the upper fraction cB to the distribution bit number B of the consumed bit number c, (T) of the quantized normalized coefficient from which the corresponding sign has not been removed, which is the remainder obtained by removing the quantized normalized coefficient series output from the quantization normalized coefficient series outputted from the quantization normalized coefficient sequence output unit 105. [ An example of the discard code is that the number of bits corresponding to one or more quantized normalized coefficients in the region including the highest frequency is not less than c-B and is a minimum code. In other words, by setting only the quantized normalized coefficients on the low frequency side to be encoded and the remaining quantized normalized coefficients on the high frequency side as the encoding targets, the length of the corresponding variable length code is equal to or smaller than the distribution bit number B And the maximum number of samples of the quantized normalized coefficient to be encoded is t.

&Lt; Gain lower limit setting unit 108 >

When the number of consumed bits c is larger than the number of allocated bits B, the gain lower limit setting unit 108 sets the value of the present gain g (the gain g corresponding to the consumed bit number c) ) Is set as the lower limit value (g _min ) of the gain (g _min? G). The lower limit value ( _gmin ) of this gain means that the value of the gain should be at least above this value.

&Lt; First branch portion 109 >

If the gain lower limit setting unit 108, the lower limit value (g _min) of a gain is set in the first branch 109, the first gain updating unit 110. If the upper limit value (g _max), the gain is already set And if not, controls the gain expanding unit 111 to perform the next process.

<First Gain Updating Unit 110>

The first gain updating unit 110 sets a value between the present value of the gain g (the gain g corresponding to the number of consumed bits c) and the upper limit value g _max of the gain to the new Value. This is because the value of the optimum gain exists between the value of the current gain g and the upper limit value g _max of the gain. First gain update unit 110, for example, is set as the gain (g) a new mean value of the gain (g) values and the gain upper limit value (g _max) of the current time (g ← (g + g _max) /2). The value of the gain (g) for this time is also that it is set as a lower limit value (g _min) of the gain, the new mean value of the upper limit value (g _max) and the gain lower limit value (g _min) of the gain set to the value of the gain (g) (G? ( _Gmax + _gmin ) / 2). And then returns to the processing of the quantization unit 105. [

<Gain Expansion Unit 111>

The gain enlarging unit 111 obtains the number of samples (s) from the quantized normalized coefficient on the lowest frequency side to the quantized normalized coefficient on the highest frequency side where the value is not 0, The larger the value (u = st) obtained by subtracting the number of samples t, the larger the increment from the current gain to the new gain. For example, the new gain (g) ← the gain (g) of this time × (1 + u / N × α). Here, α is a predetermined constant.

Alternatively, the gain expanding unit 111 may increase the gain from the present gain to a new gain as v = Nt obtained by subtracting the number of samples (t) outputted by the number-of-samples measuring unit 118 from the number of samples N to be encoded . For example, the new gain (g) ← the gain (g) of this time × (1 + v / N × α).

That is, the gain enlarging unit 111 increases the value of the gain g (g) as the value obtained by subtracting the number of samples of the quantized normalized coefficient from which a corresponding code is not removed from a part or all of the number of samples of the quantized normalized sample sequence, . And then returns to the processing of the quantization unit 105. [ In other words, the gain enlarging unit 111 increases the value obtained by subtracting the number of samples of the quantized normalized coefficient from which a corresponding code is not removed from a part or all the sample numbers of the quantized normalized sample sequence, The value of the gain is updated so that the increment from the value to the updated value becomes larger, and the subsequent processing of the quantization unit 105 is performed.

<Gain Reduction Updating Unit 132>

The gain reduction updating unit 132 sets a value (g '<g) smaller than the current gain g as a new gain. The gain reduction unit 132 includes a gain upper limit setting unit 112, a second branching unit 113, a second gain updating unit 114, and a gain reducing unit 115.

<Gain upper limit setting unit 112>

When the number of consumed bits c is smaller than the number of allocated bits B, the gain upper limit setting unit 112 sets the value of the present gain g (the gain g corresponding to the consumed bit number c) Value) is set to the upper limit value (g _max ) of the gain (g _max? G). The upper limit value ( _gmax ) of this gain means that the value of the gain should be at least below this value.

&Lt; Second branching part 113 >

When the upper limit value g _max of the gain is set in the gain upper limit setting unit 112, the second branching unit 113, if the lower limit value g _{min of the} gain is already set, the second gain updating unit 114 ), And if not, controls the gain reduction unit 115 to perform the next process.

&Lt; Second gain updating unit 114 >

Second gain updating unit 114 is the value of the gain (g) of the current time values between (consumed bit number (the value of the gain (g) corresponding to c)), and the lower limit value (g _min) of the gain the gain (g ) As the new value. This is because the value of the optimum gain is to exist between the lower limit value (g _min) of a gain value and a gain (g) of the current time. Second gain update unit 114, for example, and sets the average value of the gain (g) the lower limit value (g _min) of the value and the gain of the current time as the value of the new gain (g) (g ← (g + g _min ) / 2). The value of the gain (g) for this time is also that it is set as the upper limit value (g _max) of the gain, the new mean value of the upper limit value (g _max) and the gain lower limit value (g _min) of the gain set to the value of the gain (g) (G ← (g _max + g _min ) / 2). And then returns to the processing of the quantization unit 105. [

&Lt; Gain reducing unit 115 >

The gain reduction unit 115 reduces the gain of the new gain g from the value of the present gain g as the number of redundant bits Bc which is a value obtained by subtracting the consumption bit number c from the distribution bit number B, So that the decrease in the value is increased. However, the value of the new gain g is also a positive value. For example, the new gain (g) ← the gain (g) of this time × (1- (B-c) / B × β). Here, β is a predetermined constant. That is, the gain reduction unit 115 greatly reduces the value of the gain g as the value B-c obtained by subtracting the consumption bit number c from the distribution bit number B is larger. And then returns to the processing of the quantization unit 105. [ In other words, as the value Bc obtained by subtracting the number of consumed bits c from the number of allocated bits B from the number of allocated bits B is greater, the gain reduction unit 115 reduces the gain g from the value before update to the value after update The value of the gain g is updated so as to be larger, and the subsequent processing of the quantization unit 105 is performed.

<Discard part 116>

When the number of consumed bits c output from the determining unit 107 is larger than the number of allocated bits B, the discarding unit 116 sets the number of consumed bits c (c) out of the integer signal codes output from the determining unit 107 ) Is removed from the code corresponding to the quantized normalized coefficient on the high frequency side by the number of codes which exceeds the distribution bit number (B), as a new integer signal code. That is, the discard unit 116 stores the sign (discard code) corresponding to the quantized normalized coefficient on the high frequency side corresponding to the upper fraction cB with respect to the distribution bit number B of the consumed bit number c as the integer signal code (Sampled column code) obtained by removing the residual code (sample code) from the sample code.

&Lt; Gain coding unit 117 >

Obtains a gain code by encoding the gain output from the determining section 107 to a predetermined number of bits, and outputs the gain code.

[Modifications of First Embodiment]

&Lt; Encoder 150 >

The encoding process performed by the encoding apparatus 150 according to the modification of the first embodiment will be described with reference to FIG. The encoding apparatus 150 according to the modification of the first embodiment is different from the encoding apparatus 100 according to the first embodiment in that the estimated number of bits of the integer signal code is used as the number of consumed bits instead of the number of bits of the integer signal code obtained in the variable- (c). The encoding apparatus 150 includes a gain update loop processing unit 190 in place of the gain update loop processing unit 130 of the encoding apparatus 100. [ The gain update loop processing unit 190 includes a bit number estimation unit (not shown) in place of the variable length coding unit 106, the determination unit 107, the gain expansion update unit 131, and the shuffle unit 116 of the gain update loop processing unit 130 156, a determining section 157, a gain expansion updating section 191, and a variable length coding section 159. The gain enlarging and updating unit 191 includes a gain enlarging unit 151 and a sample number measuring unit 168 instead of the gain enlarging unit 111 and the sample number measuring unit 118 of the gain enlarging and updating unit 131. [

Only differences from the first embodiment will be described below.

&Lt; Bit number estimation unit 156 >

Bits estimation unit 156 is quantized normalized coefficient sequence (X _Q (1), ···, X _Q (N)), the estimated value of the number of bits of the code obtained by variable length coding to obtain (the number of estimated bit) output do. In the modified example of the first embodiment, this estimated bit number is called the consumed bit number c.

&Lt; Judgment section 157 >

The determining section 157 outputs the gain g and the quantized normalized coefficient series X _Q (1), ..., X _Q (N) when the number of times of updating the gain is a predetermined number.

If the number of times of updating of the gain is less than the predetermined number, the gain enlargement updating unit 191 sets the number of bits (c) when the number of consumed bits c estimated by the bit number estimating unit 156 is larger than the allocated bit number (B) When the number of consumed bits c estimated by the estimating unit 156 is smaller than the number of allocated bits B, the gain reduction updating unit 132 performs the following processing. When the number of consumed bits c estimated by the bit number estimating unit 156 is equal to the number of allocated bits B, the determining unit 157 determines the gain g, the quantized normalized coefficient sequence X _Q ( 1), ..., X _Q (N)).

&Lt; Sample number measuring section 168 >

When the number of consumed bits c is larger than the number of allocated bits B, the number-of-samples measuring section 168 sets the sign corresponding to the upper portion cB to the number of allocated bits B of the consumed bit number c (X _Q (1), ..., X _Q (N)) output from the quantization unit 105, which are the quantization normalized coefficients of the higher frequency side to which the quantization normalization coefficient (T) of the quantized normalized coefficients of the quantized coefficients.

<Gain Expansion Unit 151>

The gain expanding section 151 is configured to change the number of samples t output by the sample number measuring section 168 instead of the number of samples t output by the sample number measuring section 118 in the gain expanding section 111 of the first embodiment, Is used.

That is, the gain enlarging unit 151 obtains the number of samples (s) from the quantized normalized coefficient on the lowest frequency side to the quantized normalized coefficient on the highest frequency side where the value is not 0, The larger the value (u = st) obtained by subtracting the number of samples t output from the current gain, the larger the increment from the present gain to the new gain. For example, the new gain (g) ← the gain (g) of this time × (1 + u / N × α). Here, α is a predetermined constant.

Alternatively, the gain expanding unit 151 may increase the gain from the current gain to the new gain (V) by increasing the value v = Nt obtained by subtracting the number of samples (t) output from the sample number measurer 118 from the number of samples To be increased. For example, the new gain (g) ← the gain (g) of this time × (1 + v / N × α).

That is, the gain increasing unit 151 obtains the value of the gain g as the value obtained by subtracting the number of samples of the quantized normalized coefficient from which a corresponding code is not removed from a part or all of the number of samples of the quantized normalized sample sequence . And then returns to the processing of the quantization unit 105. [ In other words, the gain expander 151 extracts the quantized normalized coefficient on the high frequency side, which is the subject of the discard code, from a part or all of the number of samples of the quantized and normalized sample sequence by the quantization normalization the coefficient line _{(X Q (1), ···} , X Q (N)) the greater the value obtained by subtracting the number of samples of the quantized normalized coefficients of the remaining (t) has been removed from, updated from the value prior to the update of the gain after The value of the gain is updated so that the increment to the value becomes larger, and the subsequent processing of the quantization unit 105 is performed.

<Variable-length coding unit 159>

Variable length coding unit 159 to variable length coding for judging quantized normalized coefficient sequence output from the (157) (X _Q (1), ···, X _Q (N)) to obtain a code, the code obtained And outputs it as an integer signal code (sample code). When the code of the number of bits exceeding the allocation bit number B is obtained by the variable length coding, the variable length coding unit 159 stores, in the code obtained by the variable length coding, the number of bits exceeding the distribution bit number B Is removed from the code corresponding to the quantized normalized coefficient on the high frequency side, and is output as an integer signal code.

[Second Embodiment]

&Lt; Encoding device 200 >

The encoding process performed by the encoding apparatus 200 of the second embodiment will be described with reference to FIG. The encoding apparatus 200 of the second embodiment differs from the encoding apparatus 100 of the first embodiment in that a gain update loop processing unit 230 is provided in place of the gain update loop processing unit 130 and a gain update loop processing unit 230 In place of the quantization unit 105, the determination unit 107, the gain expansion update unit 131, and the discarding unit 116 of the gain update loop processing unit 130, the quantization unit 205, the determination unit 207, After the processing of the first gain updating unit 110, the second gain updating unit 114 and the gain reducing unit 115, the quantization unit 105 The process returns to the process of the quantization unit 205 instead of returning to the process of FIG. The gain increase updating unit 231 does not include the sample number counting unit 118 in the gain increase updating unit 131 of the first embodiment and includes the gain lower limit setting unit 108 and the first branching unit 109 A first gain updating unit 110, and a gain increasing unit 211. [ Only differences from the first embodiment will be described below.

<Quantization Unit 205>

The quantization unit 205 multiplies each coefficient (each sample) of weighted normalized MDCT coefficient columns X _N (1), ..., X _N (N) (sample sequences derived from input acoustic signals of a predetermined section) (g), and obtains and outputs a quantized normalized coefficient sequence (X _Q (1), ..., X _Q (N)) which is a sequence based on an integer value (a quantized normalized sample).

&Lt; Judgment section 207 >

When the number of times of updating the gain is a predetermined number, the determining section 207 outputs the gain, the integer signal code, and the number of consumed bits (c).

When the number of times of updating of the gain is less than the predetermined number of times, when the number of consumed bits c measured by the variable length coding unit 106 is larger than the number of allocated bits B, the gain enlargement updating unit 231 calculates When the number of consumed bits c measured by the encoding unit 106 is smaller than the number of allocated bits B, the gain reduction updating unit 132 controls the processing described in the first embodiment. When the number of consumed bits c measured by the variable length coding unit 106 is equal to the number of allocated bits B, the determining unit 207 outputs the gain, integer signal code, do.

<Discard part 216>

When the number of consumed bits c output from the determining unit 207 is larger than the number of allocated bits B, the discarder 216 sets the number of consecutive bits c) is removed from the code corresponding to the quantized normalized coefficient on the high frequency side by the number of codes that exceeds the distribution bit number (B), and outputs as a new integer signal code. That is, the discard unit 216 stores the sign (discard code) corresponding to the quantized normalized coefficient on the high frequency side corresponding to the upper fraction cB with respect to the distribution bit number B of the consumed bit number c as the integer signal code (Sample column code), and outputs the remaining code (sample process code subjected to the decimation process) as a new integer signal code.

<Gain Expansion Unit 211>

The gain increasing unit 211 increases the increase from the current gain to the new gain as the number of insufficient bits cB, which is a value obtained by subtracting the number of distributed bits B from the number of consumed bits c. For example, the new gain (g) ← the gain (g) of this time × (1+ (cB) / B × α). Here, α is a predetermined constant. That is, when the number of consumed bits c is larger than the number of allocated bits B and the upper limit value g _{max of the} gain is not set, The value of the gain g is greatly increased as the value cB obtained by subtracting the number B is larger. And then returns to the processing of the quantization unit 205. [ In other words, the gain enlarging unit 211 increases the gain c from the value before the update of the gain g to the value after the update g as the value cB obtained by subtracting the number of distribution bits B from the number of bits consumed c is larger, The value of the gain g is updated so that the subsequent processing of the quantization unit 205 is performed.

[Modified example of the second embodiment]

&Lt; Encoder 250 >

The encoding process performed by the encoding device 250 of the modification of the second embodiment will be described with reference to Fig. The encoding apparatus 250 of the modification of the second embodiment is different from the encoding apparatus 200 of the second embodiment in that the number of bits of the integer signal code is replaced with the number of bits of the integer signal code, (C). The encoding apparatus 250 includes a gain update loop processing unit 290 in place of the gain update loop processing unit 230 of the encoding apparatus 200. The gain update loop processing unit 290 includes a gain update loop processing unit 230, A bit number estimating unit 156, a variable length coding unit 159 and a determining unit 257 instead of the coding unit 106, the discarding unit 216 and the determining unit 207. [ Only the difference from the second embodiment will be described below.

&Lt; Bit number estimation unit 156 >

The bit number estimating unit 156 is the same as the modification of the first embodiment.

&Lt; Judgment section 257 >

The determination unit 257 outputs the gain, the quantized normalized coefficient series, and the number of consumed bits (c) when the number of update times of the gain is a predetermined number.

If the number of times of updating of the gain is less than the predetermined number, the gain enlargement updating unit 231 sets the number of bits (C) when the number of consumed bits c estimated by the bit number estimating unit 156 is larger than the distribution bit number When the number of consumed bits c estimated by the estimating unit 156 is smaller than the number of allocated bits B, the gain reduction updating unit 132 controls the processing described in the first embodiment. If the number of consumed bits c estimated by the bit number estimating unit 156 is equal to the number of allocated bits B, the determining unit 257 determines the gain, the quantized normalized coefficient sequence, the number of consumed bits c, .

<Variable-length coding unit 159>

The variable length coding unit 159 is the same as the modification of the first embodiment.

[Third embodiment]

&Lt; Encoder 300 >

The encoding process performed by the encoding apparatus 300 of the third embodiment will be described with reference to Fig. The encoding apparatus 300 of the third embodiment differs from the encoding apparatus 100 of the first embodiment in that the gain lower limit setting unit 108, the first gain updating unit 110, the gain upper limit setting unit 112, The first gain updating unit 310, the gain upper limit setting unit 312, the second gain updating unit 314 and the consumption bit number storage unit 320 . The gain enhancement updating unit 331 includes a gain lower limit setting unit 108 of the gain enhancement updating unit 131 and a gain lower limit setting unit 308 and a first gain updating unit 310 in place of the first gain updating unit 110. [ Respectively. The gain reduction unit 332 includes a gain upper limit setting unit 312 and a second gain updating unit 314 instead of the gain upper limit setting unit 112 and the second gain updating unit 114 of the gain reduction updating unit 132, Respectively. The gain update loop processing unit 330 includes a gain increase update unit 331 and a gain reduction update unit 332 instead of the gain increase update unit 131 and the gain reduction update unit 132 of the gain update loop processing unit 130 do. Only differences from the first embodiment will be described below.

&Lt; Gain lower limit setting unit 308 >

Gain lower limit setting unit 308 sets the value of the gain (g) of this time as a lower limit value (g _min) of the gain (g _min ← g). The gain lower limit setting unit 308 stores the number of consumed bits c in the number of consumed bits storage unit 320 as the number of consumed bits c _L at the time of setting the lower limit. That is, when the number of consumed bits c is larger than the number of allocated bits B, the gain lower limit setting unit 308 sets the gain lower limit setting unit 108 in accordance with the processing of the lower gain setting unit 108 of the first embodiment, c is set as the number of consumed bits (c _L ) in the lower limit setting and stored in the number of consumed bits storage unit 320.

<Gain upper limit setting unit 312>

The gain upper limit setting unit 312 sets the value of the current gain g to the upper limit value g _max of the gain (g _max? G). Also stored in the gain upper limit setting unit 312 sets the upper limit number of bits spent in the number of spent bits (c) (c _U), the number of bits consumed storage unit 320 as. That is, the gain upper limit setting unit 312 sets the gain upper limit setting unit 312 to the gain upper limit setting unit 112 in the case where the consumed bit number c is smaller than the distribution bit number B, c is set as the number of consumed bits (c _U ) in the upper limit setting and is stored in the number of consumed bits storage unit 320.

<First Gain Updating Unit 310>

When the number of bits consumed c is larger than the number of allocation bits B and the upper limit value g _max of the gain is already set, the first gain updating unit 310 sets the upper limit when consumed bit (c _U) to the number of bits spent in the lower limit set based on (c _L), at least one of an indicator of the possibility of a lower limit value (g _min) the upper limit value (g _max) of the indicator and the gain of the potential of the gain . The &quot; possibility index &quot; means an index indicating the possibility of the value of the gain g.

[Indicator of possibility of lower limit value (g _min ) of gain]

First gain update unit 310, for example, obtained by an indicator (w) represents the relative likelihood of the lower limit value (g _min) of the gain A in formula.

_{w = (Bc U) / (} c L -c U) ( formula A)

Expression A is semantically based on the difference between the number of allocated bits B and the number of consumed bits c _U in the upper limit setting and the difference between the number of consumed bits c _{L in the} lower limit setting and the number of allocated bits B The right side of expression B is transformed.

_{w = (Bc U) / (} Bc U + c L -B) ( formula B)

Therefore, the index w may be obtained by the expression B instead of the expression A.

Formula A or formula is at larger surface (w) as determined by the B, the lower limit value (g _min) of the gain side there is a possibility as the value of the gain (g), when the smaller the index (w), the upper limit value of the gain (g _max becomes a value of the gain g.

[Indicator of probability of upper limit value ( _gmax ) of gain]

The relative likelihood of the upper limit of the gain ( _gmax ) is (1-w).

That is, instead of obtaining the index w by the formula A or B, the index 1-w of the possibility of the upper limit value g _max of the gain may be obtained by the formula C.

(1-w) = (c L -B) / (c L -c U) ( type C)

Equation C is a difference between the difference Bc _U between the number of allocated bits B and the number of consumed bits c _U at the upper limit setting and the number of consumed bits c _L and the allocated bit number B at the lower limit setting, (c _L -B).

_{1-w = (c L -B} ) / (Bc U + c L -B) ( formula D)

Therefore, the index (1-w) may be obtained by the equation (D) instead of the equation (C).

When the index 1-w obtained by the formula A or the equation B is large, the upper limit value g _max of the gain is likely to be the value of the gain g, and when the index 1-w is small, the side of the lower limit value (g _min) is possible as the potential value of the gain (g).

Then, the first gain updating unit 310 sets a weighted average based on the larger one of the upper limit value g _max of the gain and the lower limit value g _min of the gain as the value of the new gain g outputs _{(g ← g min × w +} g max × (1-w)). That is, when the difference between the number of distributed bits B and the number of consumed bits c _U in the upper limit setting is larger than the difference between the number of consumed bits c _L and the number of allocated bits B in the lower limit setting, _min , which is close to the value of the desired gain g.

Or the first gain updating unit 310 is w = (Bc + _U C) with a constant (C) which is the value of the positive _{/ (c L -c U + 2} × C) surface (w) that the relaxation weighted as . In this case,

(1-w) = (c L -B + C) / (c L -c U + 2 × C)

And the value of the new gain g is the middle of the weighted average based on the arithmetic mean value of the upper limit value g _max of the gain and the lower limit value g _min of the gain and the difference between the number of consumed bits and the number of allocated bits.

When the number of samples of the quantized normalized sample (the number of discarded samples Tr) to be subjected to the discard code is obtained in the sample number counting section 118, the number of consumed bits c _L and the distribution bit It is also possible to use the discarded number of samples Tr instead of the difference of the number (B). This is because the larger the difference between the number of consumed bits (c _L ) and the number of allocated bits (B) in the lower limit setting, the larger the number of discarded samples Tr is. The relationship between the number of consumed bits (c _L ) and the number of allocated bits (B) in the lower limit setting and the number of discarded samples Tr is experimentally obtained in advance to estimate the number of discarded samples Tr at the lower limit setting c _L ) and the number of distribution bits (B). γ is an empirically determined coefficient for conversion and can be expressed as w = (Bc _U ) / (Bc _U + γ × Tr) by substituting (c _L -B) = γ × Tr. Similarly, also in the amount of w = constant using the values _{(C) (Bc U + C} ) / (Bc U + γ × Tr + 2 × C) to a surface (w) that the relaxation weighted. That is, the first gain updating unit 310 calculates an index of the possibility of the lower limit value of the gain and an upper limit value of the gain ( _CU ) by using the number of allocated bits B, the number of discarded samples Tr, Or an indicator of the likelihood of a failure. Although it is preferable to use the most recent number of samples Tr obtained in the processing of the latest number of samples counting section 118, it is also possible to use the number of samples Tr obtained in the processing of the past number of samples counting section 118 .

And then returns to the processing of the quantization unit 105. [

&Lt; Second gain updating unit 314 >

Less than the consumption of Bits (c) allocation of bits (B), In addition, if the lower limit value of the gain (g _min) is already set, the second gain updating unit 314 is the first gain updating unit 310, and The same operation is performed.

"Index of possibility" is the one, in which direction of the lower limit value (g _min) of a gain or upper limit value (g _max), by moving the value of a certain gain (g) indicates how close to the value of the proper gain (g). In the present embodiment, the number of updates until the gain g converges to an appropriate value can be reduced in order to update to a new value of the gain g based on this index.

Also, update the shape of the first gain updating unit 310 and the second gain unit 314 is at least one of an indicator of the possibility of a lower limit value (g _min), the upper limit value (g _max) of the indicator and the gain of the potential of the gain to obtain, on the side with the possibility of the lower limit value (g _min) and gain upper limit value (g _max) for the gain given the greater weight, gain the weighted average of the lower limit value (g _min) and gain upper limit value (g _max) for the gain (g) With a new value of. However, the first gain updating unit 310 and the second gain update unit 314 does not produce the possibility of indicators of potential side to give the largest weight that of the lower limit value (g _min) and gain upper limit value (g _max) of the gain , it may be a weighted average of the lower limit value (g _min) and upper limit value gain (g _max) of the gain to the new value of the gain (g). For example, the first gain updating unit 310 and the second gain updating unit 314 may obtain neither the index w nor (1-w), but the consumption bit number c _U and the lower limit setting On the basis of the number of consumed bits (c _L ) and the number of allocated bits (B)

(Equation 3)

or,

(Equation 4)

May be obtained as a new value of the gain g. That is, the larger the difference between the number of distributed bits B and the number of consumed bits c _{U in} the upper limit setting is, the larger the weight of the upper limit value g _max of the gain is given, or the number of consumed bits c _L and the distribution When the weighted average of the number of bits (B) the car, on the side lower limit value (g _min) of a gain given the greater weight, the lower limit of the gain (g _min) and upper limit value (g _max) of the gain larger of the new value of the gain (g) And there is no limit to the process.

Alternatively, when the first gain updating unit 310 and the second gain updating unit 314 are configured to update the gain g based on the number of discarded samples Tr, end

(Equation 5)

or,

(Equation 6)

May be obtained as a new value of the gain g.

Also, for example, the lower limit value (g _min) and gain upper limit value (g _max) to give the of the weight to any one of the gain the weighted average of the lower limit value (g _min) and gain upper limit value (g _max) for the gain of the gain ( g). For example,

(? ₁ x _gmin + _gmax ) / (? ₁ +1)

May be a new value of the gain g. Here, ω ₁ is, for example, g _min p. In this case there is the potential, that is, (Bc _U)> In the case of (c _L -B) to take a value of at least one quantity, g _max side is possibly (B c _U ) <(c _L -B), a positive value of 1 or less may be taken, and a larger value may be set so as to take a larger value as B c _U is larger. For example, ω ₁ may be a monotone increasing function value related to Bc _U. or,

* (g _min + w ₂ x g _max ) / (1 + w ₂ )

May be a new value of the gain g. Here, ω ₂ takes a positive value of, for example, g _max if there is a possibility, and takes a positive value of 1 or less if g _min is more likely, and c _L -B The larger the value, the larger the value is set. For example, ω ₂ may be a monotone increasing function value related to c _L -B. Alternatively, supposing that? ₃ is a positive value of 1 or more and a monotone increasing function value relating to Bc _U is taken, and? ₄ is a positive value of 1 or more, taking a monotone increasing function value with respect to c _L -B, g _min is more likely ((Bc _U )> (c _L -B))

(? ₃ x _gmin + _gmax ) / (? ₃ +1)

For the gain (g) if there is a new value, g _max p is a possibility of ((for _{Bc U) <(c L -B} ))

(g _min +? ₄ x g _max ) / (1 +? ₄ )

May be a new value of the gain g.

Thus, the weight based on at least the number of bits distributed B, the number of consumed bits c _L in the lower limit setting and the number of consumed bits c _{U in the} upper limit setting is set to the upper limit value g _max of the gain and the lower limit value of the gain _gmin ) may be a weighted average of the upper limit value of the gain and the lower limit value of the gain as the updated gain.

[Modifications of Third Embodiment]

The third embodiment is different from the first embodiment in that the gain lower limit setting section 108, the gain upper limit setting section 112, the first gain updating section 110 and the second gain updating section 114 are replaced The gain lower limit setting unit 108, the gain upper limit setting unit 112, the first gain updating unit 110 and the second gain updating unit 114 of the second embodiment are the same as those described in the third embodiment And the gain lower limit setting unit 1008, the gain upper limit setting unit 1012, the first gain updating unit 1010, the second gain setting unit 1010, and the second gain setting unit 1010 of the TCX encoding apparatus 1000 described in [ The updating unit 1014 may be replaced with the one described in the third embodiment.

Alternatively, the gain lower limit setting unit 108, the gain upper limit setting unit 112, the first gain updating unit 110, and the second gain updating unit 114 of the modified example of the first embodiment may be the same as those in the third embodiment The gain lower limit setting unit 108, the gain upper limit setting unit 112, the first gain updating unit 110 and the second gain updating unit 114 of the modified example of the second embodiment may be replaced with the above- May be replaced with those described in the third embodiment of Figs.

That is, when the number of bits of the code or the number of bits to be obtained by coding a column of the integer value samples obtained by dividing each sample of the sample string by the gain before the update is larger than the predetermined number of distribution bits (B) the lower limit value (g _min) settings, and set the number of bits or the estimated number of bits as the (c _L) the number of bits spent during the minimum set as a, and updates the respective samples sample columns by dividing by the gain before the heat by the integer value of the sample obtained The gain before the update is set as the upper limit value ( _gmax ) of the gain and the number of bits or the estimated bit number is set as the upper limit when the number of bits or the number of bits to be obtained by encoding is smaller than the predetermined number set as the number of bits _(U c), and the allocated number of bits (B) and the lower limit set at the bit number of consumption _(L c) to the number of bits spent during the maximum set _(U c) at least based on the upper limit of the weight gain of the (g _max) assigned to at least one of the gain and the lower limit value (g _min) of a, it may be a weighted average of the gain of the upper and lower limits of the gain in the gain after update.

<Hardware Configuration Example of Encoding Apparatus>

The encoding apparatus according to the embodiment includes an input section to which a keyboard or the like can be connected, an output section to which a liquid crystal monitor or the like can be connected, a CPU (Central Processing Unit) ), A ROM (Read Only Memory), an external storage device such as a hard disk, and an input / output section of these, a CPU, a RAM, a ROM, and an external storage device. If necessary, a device (drive) capable of reading and writing a storage medium such as a CD-ROM may be provided in the encoding apparatus.

The external storage device of the encoding apparatus stores a program for performing encoding and data required for processing of the program and the like (not limited to an external storage device, and for example, a program may be stored in a ROM It is also possible to memorize it. The data obtained by the processing of these programs are appropriately stored in the RAM or the external storage device. Hereinafter, a storage device for storing data or an address of the storage area is simply referred to as a &quot; storage unit &quot;. A program for executing encoding is stored in the storage unit of the encoding apparatus.

In the encoding apparatus, each program stored in the storage unit and data necessary for the processing of each program are read into the RAM as needed, and analyzed and processed by the CPU. As a result, coding is realized by the CPU realizing a predetermined function.

<Supplementary information>

The present invention is not limited to the above-described embodiment, but can be appropriately changed without departing from the gist of the present invention. For example, in each of the above-described embodiments, when the number of consumed bits is smaller than the number of allocated bits, the processing of the gain reduction updating unit is performed. When the number of consumed bits equals the number of allocated bits, the determination unit outputs a gain or the like. However, the processing of the gain reduction updating unit may be performed when the number of consumed bits is not larger than the allocation bit number. The processes described in the above embodiments may be executed not only in time series in accordance with the description order but also in parallel or individually depending on the processing capability of the apparatus for executing the processing or the necessity.

In the case where the processing function in the hardware entity (encoding apparatus) described in the above embodiments is realized by a computer, the processing contents of the functions that the hardware entity should have are described by the program. By executing this program on a computer, a processing function in the hardware entity is realized on a computer.

The program describing the processing contents can be recorded on a computer-readable recording medium. An example of a computer-readable recording medium is a non-transitory recording medium. As a computer-readable recording medium, for example, a magnetic recording apparatus, an optical disk, a magneto-optical recording medium, a semiconductor memory, or the like may be used. Specifically, for example, a hard disk device, a flexible disk, a magnetic tape, or the like can be used as the magnetic recording device, a DVD (Digital Versatile Disc), a DVD-RAM (Random Access Memory), a CD- A magneto-optical disc (MO) as a magneto-optical recording medium, an EEP-ROM (Electronically Erasable and Programmable Read-Only Memory) as a semiconductor memory, and the like Can be used.

The distribution of this program is performed by selling, transferring, lending, or the like, a portable recording medium such as a DVD or a CD-ROM recording the program, for example. The program may be stored in a storage device of a server computer, and the program may be transferred from the server computer to another computer through a network to distribute the program.

A computer that executes such a program stores, for example, a program recorded on a portable recording medium or a program transmitted from a server computer once in its own storage device. At the time of executing the processing, the computer reads the program stored in its own recording medium and executes processing according to the read program. Further, as another execution form of the program, the computer may read the program directly from the portable recording medium and execute processing according to the program. Further, each time a program is transmitted from the server computer to the computer, , And the processing according to the received program may be executed. In addition, a configuration in which the above-described processing is executed by a so-called ASP (Application Service Provider) type service that realizes a processing function only by execution instruction and result acquisition without transferring the program from the server computer to the computer You can. Incidentally, the program in this embodiment includes information to be provided for processing by the electronic calculator and includes a program (data not having a direct instruction to the computer, but having a property of specifying the processing of the computer, etc.) .

In this embodiment, a hardware entity is configured by executing a predetermined program on a computer. However, at least a part of the processing contents may be implemented in hardware.

100, 150, 200, 250, 300, 1000 encoding apparatus

Claims (3)

A method for encoding a sample sequence derived from an input acoustic signal of a predetermined section,
A quantization step of quantizing a value obtained by dividing each sample of the sample train by a gain to obtain a quantized normalized sample train;
A gain enlargement updating step of setting a value larger than the gain as a new gain;
A gain reduction updating step of setting a value smaller than the gain as a new gain;
Wherein when the number of update times of the gain is a predetermined number of times, a variable length coding step is performed, and when the number of times of updating of the gain is less than a predetermined number and the number of consumed bits, which is the estimated bit number of the code corresponding to the quantized normalized sample string, The gain enlarging and updating step is carried out when the number of times of updating of the gain is larger than the predetermined number and the number of bits consumed is smaller than the predetermined number of allocated bits, Step,
The gain enlarging and updating step includes:
A gain lower limit value setting step of setting a gain value corresponding to the consumed bit number as a lower limit value of the gain when the consumed bit number is larger than the allocation bit number;
The larger the value obtained by subtracting the number of allocation bits from the number of consumption bits in the case where the consumption bit number is larger than the allocation bit number and the upper limit value of the gain is not set, A gain enlarging step of updating the value of the gain so as to increase the increment to the quantization step; And
Wherein when the number of consumed bits is larger than the number of allocated bits and the upper limit of the gain is already set, the number of allocated bits (B) and the number of consumed bits, which is the number of consumed bits, The number of consume bits in the setting (c _L ), the number of consumed bits in the upper limit setting (c _U ), the lower limit value (g _min ) of the gain, which is the number of consumed bits when the consumed bit number is smaller than the allocation bit number, The upper bound (g _max ) and the positive constant (C)
(Equation 11)

As a new value of the gain;
Lt; / RTI &gt;
The gain reduction step includes:
A gain upper limit value setting step of setting a gain value corresponding to the consumed bit number as an upper limit value of gain when the consumed bit number is smaller than the distribution bit number;
Wherein the gain is determined such that the larger the value obtained by subtracting the consumed bit number from the distribution bit number when the consumption bit number is smaller than the distribution bit number and the lower limit value of the gain is not set, A gain reduction step of updating the value of the gain so that the reduction to the value becomes larger and causing the quantization step to be performed; And
When the consumption bit number is smaller than the distribution bit number and the lower limit value of the gain is already set,
(Equation 12)

As a new value of the gain;
Lt; / RTI &gt;
Wherein the variable length coding step performs variable length coding on the quantized normalized sample sequence to obtain a sample sequence code. An apparatus for encoding a sample sequence derived from an input acoustic signal of a predetermined section,
A quantization unit for quantizing a value obtained by dividing each sample of the sample train by a gain to obtain a quantized normalized sample train;
A gain enlargement updating unit that sets a value larger than the gain as a new gain;
A gain reduction updating unit that sets a value smaller than the gain as a new gain; And
And when the number of update times of the gain is a predetermined number of times, a process of the variable length coding unit is performed, and when the number of times of updating of the gain is less than a predetermined number and the number of bits consumed, which is the estimated bit number of the code corresponding to the quantized normalized sample sequence The processing of the gain enlargement updating unit is performed when the number of times of updating of the gain is less than a predetermined number of times and the number of consumed bits is smaller than the number of allocation bits, And a judgment unit for judging,
The gain enlargement /
A gain lower limit value setting unit which sets a gain value corresponding to the consumed bit number as a lower limit value of the gain when the consumed bit number is larger than the allocation bit number;
The larger the value obtained by subtracting the number of allocation bits from the number of consumption bits in the case where the consumption bit number is larger than the allocation bit number and the upper limit value of the gain is not set, A gain enlarging unit for updating the value of the gain so as to increase the increment to the quantization unit; And
Wherein when the number of consumed bits is larger than the number of allocated bits and the upper limit of the gain is already set, the number of allocated bits (B) and the number of consumed bits, which is the number of consumed bits, The number of consume bits in the setting (c _L ), the number of consumed bits in the upper limit setting (c _U ), the lower limit value (g _min ) of the gain, which is the number of consumed bits when the consumed bit number is smaller than the allocation bit number, The upper bound (g _max ) and the positive constant (C)
(Expression 21)

A first gain updating unit for setting a gain of the gain control unit as a new value of the gain;
Lt; / RTI &gt;
Wherein the gain reduction updating unit includes:
A gain upper limit value setting unit for setting a gain value corresponding to the consumed bit number as an upper limit value of the gain when the consumed bit number is smaller than the distribution bit number;
Wherein when the consumption bit number is smaller than the distribution bit number and the lower limit value of the gain is not set, the larger the value obtained by subtracting the consumption bit number from the distribution bit number, the more updated A gain reduction unit that updates the value of the gain so as to increase the decrease to a later value and causes the quantization unit to perform processing; And
When the consumption bit number is smaller than the distribution bit number and the lower limit value of the gain is already set,
(Equation 22)

A second gain updating unit that sets the new gain value as a new value of the gain;
Lt; / RTI &gt;
Wherein the variable length coding unit performs variable length coding on the quantized normalized sample sequence to obtain a sample sequence code. A computer-readable recording medium storing a program for causing a computer to execute the steps of the encoding method of claim 1.

Images